Foldit Drug Design Blog: Interface Update
It's time for another update on the process we have been using to update the graphics for Foldit drug design!
This project started awhile ago with a very simple interface.
The idea that we had was to use the graphical user interface for designing proteins. In the image above, you see that we have a pi-menu pop up when you select an atom to design. While this is a pretty cool concept, the main problem that we encountered was that the pi-menu hid the ligand that you were supposed to be designing. It was very difficult to perform modifications and see how it effected the protein/ligand.
This difficulty resulted in a move to a separate menu.
In this menu, which we call the Ligand Design Panel, you can select atoms to design and then click on the design panel to change those atoms. The cool thing is now you can move the design panel around and view the protein and the ligand without interference with the menu. There are of course, some problems with this menu. We have some mixed opinions on this, and invite you to share your thoughts in our thread below.
The elements that you can choose from are just labeled with the element name: C-carbon, N-nitrogen, O-oxygen, P-phosphorus, etc. Also, the fragments shown below the elements, are kind of an ugly magenta. Further, when you click on an atom, then a fragment, you have no idea where that new fragment will be placed, spatially. So, we needed to update the ligand panel.
The new ligand panel is much more colorful! The elements are colored based on their CPK coloring and the fragments have been replaced with high-resolution images. Further, now, when you select a fragment, a glowing outline of that fragment is drawn onto the structure! No more guessing where a new fragment will be placed.
Additionally, we added the ability to modify bonds. Which brings us to the next graphical improvement. Before, everything was shown as a single bond. Now, the small molecule is drawn with its bonds shown and a new way of viewing the protein, called the Cartoon Ligand view option (under advance settings, View Protein: Cartoon Ligand) is available.
We have added new ways of viewing interactions within the protein.
We have also added a ligand viewing panel, which allows for turning the isosurface on only around the small molecule, showing areas where there are a hydrogen bond acceptor/donor, and finally, where there is repulsion between atoms. You can also change the transparency of everything on the fly, which allows you more direct manipulation of the settings.
Finally, we have added functionality that helps notify you if you are trying to design something that is not chemically feasible.
We hope everyone enjoys all the new additions. Let us know what other things you would like to see in the graphics department.( Posted by free_radical 82 2411 | Tue, 03/24/2015 - 15:50 | 11 comments )
Through the eyes of a scientist: Part 2 - Puzzle 1052
In addition to highlighting some of our favorite Scientist-Shared solutions, we'll look at several lower-scoring designs to illustrate issues relating to:
- hydrogen bond networks
- buried, unsatisfied polar atoms
Check it out and leave your questions in the comments below!( Posted by bkoep 82 1179 | Sat, 02/28/2015 - 00:25 | 12 comments )
Player designs enter the wet lab
Last week the Baker Lab ordered materials to construct the latest batch of Foldit designs in the lab for experimental testing! The following eight protein designs were selected based on visual inspection by our scientists and folding predictions by the Rosetta@home distributed computing project. For each design below, we've included an image of the Foldit player's design on the left; and on the right, a folding funnel with the energy and Cα-RMSD for 100000s of Rosetta@home predictions in red, as well as an image of the lowest-energy prediction. We like to see that the lowest-energy prediction also has a low RMSD (explained here).
For more information about the types of experiments in store for these Foldit player-designed proteins, see our previous blog post.
In monomer designs, we were looking particularly for diverse topologies containing some β-sheet secondary structure. Such topologies are significantly more difficult to design than the helical bundles that have been so successful in the past. Consequently, these folding funnels may appear less pristine, but we are still very excited to experiment with them in the lab!
Symmetric Oligomer Designs
For designs of symmetric homooligomers, we do a similar analysis to make sure the monomer will fold up as expected.
We also want to make sure that, once folded, the monomer units are likely to bind to each other in the correct orientation. The most common reason a design fails this "docking" test is that its interface is completely hydrophobic and featureless. In such a case, two properly folded monomers can usually come together in a number of different ways to bury the same amount of hydrophobic surface. The best way to ensure specific binding in the correct orientation is to design rugged, complementary surfaces (e.g. large interdigitated side chains) and incorporate hydrogen bond networks at the interface.
The following symmetric designs performed decently for both monomer folding and docking tests (data not shown), and we are excited to try them out—however, we think there is room for improvement in the design of specific interfaces!
The players responsible for the above designs certainly deserve recognition, but there are many, many more exciting designs that just missed the cut or are still under analysis. We can't wait to see what Foldit players come up with next—keep up the great folding!( Posted by bkoep 82 1179 | Tue, 02/10/2015 - 15:57 | 9 comments )
Foldit Drug Design Part Two
My name is Sandeep Kothiwale (aka fragmentor). I am continuing the Foldit drug design blog this week. I am a graduate student at Vanderbilt University and developing the drug design module of Foldit. This blog describes the shake/wiggle feature for small molecules which is analogous to the one for protein molecule.
Drug molecules (small molecules) bind to a target molecule (protein in our case) and effect the function of the protein. This change in protein function leads to the desired physiological effect of relieving disease or its symptoms. For example, Imatinib (Gleevec) binds and blocks an enzyme whose over-activity causes leukemia.
As with imatinib, all drug molecules bind their targets in a specific pocket in a particular 3D arrangement. For successful drug design, one needs to recapitulate the expected binding pose of the putative drug (ligand) to the protein. This requires that 3D structure of ligand be determined which is able to bind the target. Spatial arrangement that atoms in a molecule can adopt with respect to each other is called a conformation. A molecule can adopt multiple freely convertible conformations by rotations about individual single bonds. Thus enumeration of 3D conformations is essential in modeling ligand binding in Foldit. As you might know, we use the wiggle feature for enumerating side-chain conformations. This is accomplished using a set of rules that have been identified for 20 or so amino acids from known protein structures in the Protein Data Bank (PDB). As you can imagine enumeration of small molecule conformation is substantially more complex than wiggle for 20 or so amino acid side chains because of large chemical space.
Foldit will use an algorithm that I helped develop for sampling conformations of ligands. It uses information contained in the Cambridge Structure Database (CSD), a repository of small molecule crystal structures (on a side note, the CSD group has let us use their database free of charge!). The algorithm uses a CSD-derived database, the csd-rotamer library that contains statistics about most commonly seen conformations of small molecular fragments. Given a molecule of interest, the algorithm determines which smaller fragments are part of it and uses information in the csd-rotamer library to sample conformations.
During the drug-design process ligand will be built by adding fragments to the base fragments. One could hit the ligand-wiggle button to sample conformations of ligands and let Rosetta (Foldit’s engine) choose the conformation that best fits the binding pocket. We have a video of this cool technology above (and at the link). The video first shows adding a fragment to the base small molecule (shown in orange) and then at 26s, the new fragment rotates. We are using HIV protease as a test case. Check it out!( Posted by fragmentor 82 2411 | Mon, 02/02/2015 - 18:14 | 4 comments )
It's been several months since CASP11 ended in August of 2014, and in December assessors presented their full analysis of CASP predictions at the CASP11 meeting in Mexico. You can see results for all teams and assessor presentations on the CASP website, but we'd like to focus for a minute on Foldit's performance.
There were other "would-be" blue ribbons, in which Foldit players produced first-rate models that we failed to select as our best. Although they did not select it as their top model, the GoScience team submitted the best overall prediction for target TR769! Likewise, WeFold used a Foldit model to develop the best overall prediction for target TR837!
Overall, we were outperformed most notably in the Refinement category by the FEIG team, which uses a vast amount of supercomputing power to explicitly simulate protein dynamics; and in the Contact-Assisted category by the LEE team, which was able to take advantage of "ambiguous contacts" that were not addressed in our Foldit puzzles.
Note: These rankings are calculated using GDT_TS, which is just one metric for evaluating model quality. The CASP website explains some other metrics that might be used to evaluate models.
With a bit of closer examination, we've concluded our troubles in the Refinement category can be divided into several distinct cases:
In the first case, Foldit players were very good at exploring solutions close to the native, but the solutions scored poorly and were not submitted as CASP predictions. Looking at the energy plot for target TR769 below, we can see that players found solutions with GDT_TS as high as 0.94(!), although their energies were less favorable than solutions near the starting model.
Energy plot for TR769. Every red dot represents a Foldit solution plotted against GDT_TS (where a value closer to 1.0 indicates a model closer to the native structure) and Rosetta energy (where a very negative value corresponds to a very high Foldit score). The blue dots represent the five solutions that were submitted as CASP predictions by the FOLDIT team. The vertical black bar represents the GDT_TS of the starting model.
In the second case, which was more common, the "energy funnel" looked good—meaning that models with better GDT_TS had more favorable energies—but Foldit players simply weren't able to explore solutions close to the native structure. In the energy plot for target TR782, for example, we can see that the best-scoring solutions at the bottom of the funnel were also the most similar to the native. Unfortunately, most Foldit players tended to move the protein away from the native conformation.
Energy plot for TR782. Every red dot represents a Foldit solution plotted against GDT_TS (where a value closer to 1.0 indicates a model closer to the native structure) and Rosetta energy (where a very negative value corresponds to a very high Foldit score). The blue dots represent the five solutions that were submitted as CASP predictions by the FOLDIT team. The vertical black bar represents the GDT_TS of the starting model.
Lastly, we saw a few targets that were both difficult to explore and difficult to score. For target TR803, solutions appeared to score better and better as they diverged from the native structure, and most Foldit players spent time moving away from the native.
Energy plot for TR803. Every red dot represents a Foldit solution plotted against GDT_TS (where a value closer to 1.0 indicates a model closer to the native structure) and Rosetta energy (where a very negative value corresponds to a very high Foldit score). The blue dots represent the five solutions that were submitted as CASP predictions by the FOLDIT team. The vertical black bar represents the GDT_TS of the starting model.
In the Contact-Assisted category, we were happy to find that Foldit players could use predicted contacts to make huge improvements in their solutions. In most cases, we posted an initial "Ts" puzzle with a limited set of simulated contacts, and then followed it up with a more complete set of "Tc" contacts. In every instance, more contacts resulted in better predictions.
For example, compare Foldit solutions for Ts/Tc827 with T0827, which was posted without contacts under the guise of 1005: De-novo Freestyle 44. Not only did additional contacts result in further exploration toward the native structure, but the complete contacts also reshaped the energy funnel to strongly favor solutions closer to the native!
Energy plot for T0827, Ts827, Tc827. Every red dot represents a Foldit solution plotted against GDT_TS (where a value closer to 1.0 indicates a model closer to the native structure) and Rosetta energy (where a very negative value corresponds to a very high Foldit score). The blue dots represent the five solutions that were submitted as CASP predictions by the FOLDIT team. The vertical black bar represents the GDT_TS of the starting model.
In the future, we'll be working to see how we can improve scoring in cases like TR769, and how to encourage more exploration for targets like TR782. We're encouraged by the progress Foldit players have made in the use of predicted contacts, and are looking forward to applying this method in future non-CASP efforts. A big thanks to all of our players for their tireless contribution to structural biology research!bkoep 82 1179 | Wed, 01/28/2015 - 01:43 | 4 comments )